"Due to", the exceptional case handled in PostProcessor

[kwalcock]

The training data generally assigns "JJ TO" as parts of speech for "due to". If for some reason we want it to be "IN TO", then an exception will need to be created. Here are some notes, though.

1. "Due to" has at least two meanings: because of and scheduled to. Here are examples of the latter:

   • The formal inauguration had been due to take place this week but was put back.
   • She had been due to fly on a TWA airliner which exploded near New York last month, but had cancelled her booking, the newspaper said.
   • Plans to do this are due to be filed in a week or so.
   • The offer, which was due to expire yesterday, is conditional on 50.1% of Dunkin' common shares.

   By treating these the same as the others, we may be seeing false causation.

2. There can be words inserted between the "due" and "to", and we are not catching these cases. Here are examples:

   • As for joint ventures, Mr. Houghton said profit was "essentially flat" due primarily to a slow recovery at Samsung-Corning Co. in Korea following a strike at a major customer and the disruption of shipments to China.
   • Apple Computer Inc. posted improved fiscal fourth-quarter profit due largely to a $48 million gain on the sale of its stock in Adobe Systems Inc.
   • "A loss of the warning symptoms of hypoglycemia is a complex problem that is very unlikely to be due simply to the type of insulin used," the American association said.
   • That is due mostly to payments from Allianz for most of the 50% stake it has agreed to acquire in Navigation Mixte's insurance business.
   • The economy's slowdown is due only partly to the austerity program launched in September 1988 to cool an overheated economy and tame inflation.
   • The increase was due mainly to a $325 million order from Turkey to equip its fleet of F-16 fighters with Loral's ALQ-178 Rapport III electronic countermeasures system.
   • But the Dow Jones Transportation Average went down for the seventh consecutive session, due largely to further selling in UAL.
   • Polish brewer Zywiec's 1996 profit slump may last into next year due in part to hefty depreciation charges, but recent high investment should help the firm defend its 10-percent market share, the firm's chief executive said.

3. Although "JJ TO" is in the majority, there are also instances of "RB TO". I do not see any difference in such sentences. The logic in the PostProcessor replaces whatever isn't IN with IN, so this doesn't matter much.

[MihaiSurdeanu]

Thanks @kwalcock !

I've assumed "due" is a preposition... If it's tagged as JJ that is fine. We just need to adjust the postprocessing rule for this.
Also, note that, theoretically, the ambiguity you mentioned ("because of" vs. "scheduled to") should be resolved through syntax, which, hopefully, our rules do. It is Ok to leave this ambiguity in at POS tag level.

Similarly, the words in between "due" and "to" should be modeled by Odin rules.

[kwalcock]

It sounds like the POS is not the important part. Using JJ might be somewhat faster because it is more likely to start like that and need less rewriting, but the cost is trivial and compatibility might be important. It must be the dependency that is the focus and the goal to get that mwe (multi-word expression?) edge in there. The training data for it looks very inconsistent. It seems like the result could be slightly improved if the mwe edge were also added if "due to" wasn't necessarily consecutive and not added if the due means scheduled. It looks like the code can't be removed, though, and that was what I was hoping for. Oh well. It's easy enough.